v2.3.7
May 13, 2026 Major releaseA big upgrade across Mock Exam, Ask a PDF, the in-reviewer tutor, and the overall look-and-feel. Highlights:
Mock Exam — built for exam prep
- Filter your exam by chapter / topic and by difficulty (Easy · Medium · Hard · All).
- Focus where it matters — optional toggles bias the sample toward cards you've struggled with in the past, and toward cards Anki considers due or overdue.
- Adaptive difficulty — opens with a warm-up, then alternates harder and easier questions so the exam feels paced rather than uniform.
- Confidence-weighted scoring — opt-in toggle. Marking yourself "confident" multiplies the reward when you're right and the penalty when you're wrong. Trains the kind of calibration high-stakes exams reward.
- Negative marking — Off, −0.25, −0.50, or −1.00 per wrong, matching common medical and pharmacy board schemes.
- Pause and resume — stop mid-exam, close Anki, come back later, and pick up exactly where you left off.
- Exam history — your last five attempts with a rolling average score, so you can see whether you're improving on a deck over time.
- Flag for review — mark questions you want to revisit, and the exam will prompt you to look at them before final submit.
- Keyboard shortcuts — number keys for multiple-choice, T/F for true-false, arrow keys to move between questions, Enter to submit. A subtle hint row reminds you which keys do what.
- Image-occlusion questions — finally supported. The masked image is shown inline, you type what you think is hidden, and the answer is graded instantly with no credit cost.
- MCQ option order is reshuffled every attempt — no more memorising "the answer is always C".
- A friendlier result page — a clean score number, a short message in your scoreband, and a quiet per-question list (collapsed by default) instead of a wall of coloured panels.
- Tag wrong answers with one click so you can review them later in the Anki browser; generate practice cards on the topics you missed; ask for a tutor explanation on any specific question with the Why? button; or export the whole result page as a PDF to share with a study partner.
- Save your exam setup as a template — keep your favourite configurations and pick one from a dropdown next time.
Smarter, more reliable AI
- The right AI model for each job — each Cardivate feature now uses a model tuned to its task. Simple classifications use a faster, cheaper model; chat and tutoring use a balanced model; card generation continues to use the strongest available. Same quality with noticeably lower credit usage on small tasks.
- Automatic failover between AI providers — if the AI service Cardivate is using has a hiccup, your request quietly retries on the other provider. Outages on either side no longer interrupt your work.
- Bulk grading for Mock Exam — free-text grading happens in one round-trip and at roughly half the credit cost of grading each question individually.
- Blank answers are free — leaving a free-text question blank no longer spends a credit; it's just marked wrong.
- Auto-tag works on scanned PDFs again — chapter detection no longer fails on high-resolution scans.
- Image-occlusion accuracy fixed — the boxes now land on the correct word for both text-based and scanned PDFs (including single-word answers).
- Friendlier error messages across every AI feature. If you're out of credits there's a top-up button; if the network is down, you're told to check your connection; transient errors show a clear retry message and confirm that no credits were charged.
Chat — Ask a PDF and the in-reviewer Tutor
- Formatted answers — replies render with italics, bold, bullet lists, numbered lists, and inline code instead of showing the formatting marks as raw text.
- Tunable look-and-feel — a new Aa Display button lets you adjust the chat's text size, font, reading width, and line spacing. Settings are remembered across sessions and shared between Ask a PDF and the Tutor.
- Three message styles — Minimal (clean), Cards (subtle tints to make "you" vs "the tutor" easy to spot at a glance), or Alternating rows.
- Calmer transcript — only a thin accent stripe and a small role label carry colour. No more highlighted backgrounds on every line.
- Remembers your last folder — both the main PDF picker and Ask a PDF open in the folder you used last time.
Image occlusion fixes
- The question now appears on the back of the card too, above the original image, so flipping a card feels symmetric.
- The red mask now lands on the correct word reliably, both on text-based and scanned PDFs.
- Chapter auto-tagging works again on scanned-PDF decks.
Update reminders are less annoying
- "Remind me later" now actually reminds you later. Previously, dismissing the update popup once silenced it permanently until the next version came out. Now it re-appears the next day if you haven't updated yet.
- Clicking Cardivate → Check for updates manually always brings the popup back up so you can re-summon it on demand.
Polish + reliability
- Card creation no longer freezes Anki during big batches — large generates feel snappy now.
- Your preferences, generation history, Mock Exam history, and downloaded OCR language packs are now preserved across add-on updates. No more re-downloading Arabic / Russian / etc. every release.
- Several small text-cropping issues fixed (the OCR Language Packs Select All / Deselect All buttons were displaying as "elect A" / "select" — now show their full labels).
- Anki versions 2.1.50 and newer are explicitly supported, and the add-on handles both old and new Anki UI toolkits.